Purchase this article with an account.
Edward Vul, Deborah Hanus, Nancy Kanwisher; Selective attention and uncertainty. Journal of Vision 2008;8(6):878. doi: 10.1167/8.6.878.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Is selective attention discrete or continuous? Are items either ‘selected’ or ‘not selected’, or might they be selected to varying degrees? If multiple items are visually selected to varying degrees, how is a response selected? Tasks used to study selective attention require the detection of co-occurrences in space or time between potential targets and a presented cue. Typically, experimenters make these tasks difficult by limiting stimulus fidelity. These manipulations increase the uncertainty inherent in the co-occurrence detection. This uncertainty may be represented on any given trial, such that a number of items are selected to varying degrees (within-trial variability) - equivalent to representing a probability distribution over likely targets. Alternatively, this uncertainty may be manifest as variability across-trials - on any trial an item is either selected or not, but which item is selected varies across trials. We asked subjects to make multiple responses on a given trial and analyzed the conditional distributions of guesses to assess the degree to which within- and across-trial variability contribute to the final distribution of reports. We assessed this for the case of temporal selective attention (RSVP task), as well as spatial selective attention (reporting a cued item from a spatial array). In both cases we find that the final distribution of reports is driven by within-trial variability. Thus, on any given trial, subjects ‘select’ a number of items to varying degrees, effectively forming a probability distribution over likely targets. Subjects then randomly sample from this probability distribution to make their responses. These results have implications for theories of selective attention - particularly, boolean map theory, as well as the phenomena of binding and crowding. Furthermore, the finding that people represent probability distributions at a given instance of time, and produce responses by sampling, has implications for cognition more broadly.
This PDF is available to Subscribers Only